At 5x, the f/8 on your lens barrel is really f/48 at the sensor. That's why your snowflake shots look smeared even when focus was perfect.
Most snowflake photographers hit the same wall. One arm of the crystal is sharp and the opposite arm is mush. You stop down to fix it and the whole frame goes soft. Then you open up and lose depth again. The depth-of-field calculators you find online don't help. They're built for portraits at two metres, not ice plates a fraction of a millimetre deep.
This article gives you the actual math: effective aperture, Airy disk size, depth of field at reproduction scale, and how much depth a tilted plate really takes up. You'll get one formula that picks the best aperture for any plate, plus tables you can use at the camera.
Why Standard DOF Calculators Fail Above 1x
Most popular depth-of-field tools assume three things:
- The subject is far away. They use focus distance and hyperfocal formulas, which stop working once magnification goes past about 0.5x.
- The marked f-number is the working f-number. It isn't. Light spreads out as the lens extends toward the subject, and the bellows factor grows quickly.
- A fixed circle of confusion of 0.03 mm. That number comes from 8×10 prints of 35 mm film. A 45-megapixel sensor has a pixel pitch of about 4.4 µm, so the old blur allowance covers almost seven pixels.
The advice built on these tools is just as stale. "f/8 is the macro sweet spot." "Diffraction kicks in at f/11." Neither statement means anything until you say what magnification you're at and what sensor you're using. f/8 at 1x behaves like f/16, and f/8 at 5x behaves like f/48. Those are different photographs.
How to Calculate Your Effective Aperture
The working f-number at the sensor is:
N_eff = N × (1 + M / P)
- N is the marked aperture.
- M is the magnification.
- P is the pupil magnification: the exit pupil diameter divided by the entrance pupil diameter.
Many symmetric macro lenses have P close to 1, which simplifies the formula to the familiar N × (M + 1). Telephoto-design macro lenses usually have P below 1, so their effective aperture is worse than the simple formula says. Reversed lenses flip the ratio.
My first winter at 5x, I shot an entire morning of sectored plates at marked f/8 because a forum thread called it the safe aperture. Every frame was soft at 100%. The camera was working at f/48, and no focusing technique fixes that.
How to Check Your Own Lens Without a Spec Sheet
Some camera and lens combinations show the effective f-number in the viewfinder, and some show the marked one. Here's a quick way to tell:
- Set manual exposure and focus the lens at its closest distance.
- Watch the displayed f-number while you rack magnification from 1x to its maximum.
- If the number climbs, the camera is reporting effective aperture. If it stays fixed, it's showing the marked value, and you have to do the math yourself.
To measure the real figure, photograph an evenly lit, defocused grey card at infinity focus and again at your working magnification. Keep shutter speed and ISO identical. The exposure difference in stops equals 2 × log₂(N_eff / N). For a lens with P = 1 at 1x, expect about a 2-stop drop.
Converting Microscope Objective NA to f-Number
If you're shooting through a finite or infinite objective, it's rated in numerical aperture rather than f-stops. Convert it with N_eff ≈ M / (2 × NA). A Mitutoyo M Plan Apo 5x has an NA of 0.14, which works out to about f/17.9 effective. That puts it right next to a conventional macro lens at f/2.8 and 5x.
How Big the Airy Disk Gets at Each Aperture
Diffraction spreads every point of light into a disk. Its diameter at the sensor is:
Airy diameter ≈ 2.44 × λ × N_eff
With green light (λ = 0.55 µm), that's about 1.34 µm per effective stop. The table below uses that wavelength and a 45 MP full-frame sensor with a 4.4 µm pixel pitch.
The DOF column needs a definition, because the old 0.03 mm circle of confusion is meaningless here. I define DOF as the depth over which defocus blur stays no larger than the Airy disk. Beyond that point, defocus becomes the dominant blur. The formula is DOF = 4.88 λ N_eff² / M².
| Marked Aperture | Magnification (M) | Effective Aperture (N_eff) | Depth of Field (µm) | Airy Disk at Sensor (µm) | Blur Spot on the Flake (µm) | Diffraction Status (45 MP) |
|---|---|---|---|---|---|---|
| f/2.8 | 2x | f/8.4 | 47 | 11.3 | 5.6 | ~2.6 px: crisp at 100% |
| f/2.8 | 5x | f/16.8 | 30 | 22.5 | 4.5 | ~5 px: soft at 100%, finest subject detail |
| f/4 | 3x | f/16 | 76 | 21.5 | 7.2 | ~5 px: soft at 100%, clean at web size |
| f/4 | 5x | f/24 | 62 | 32.2 | 6.4 | ~7 px: stacking territory |
| f/5.6 | 3x | f/22.4 | 150 | 30.1 | 10.0 | ~7 px: visible haze |
| f/8 | 4x | f/40 | 268 | 53.7 | 13.4 | ~12 px: depth with nothing sharp in it |
Look at the f/8 row. It has the most depth, and nothing inside that depth is sharp. A 268 µm zone of uniform softness isn't an achievement.
Now compare the "Blur Spot on the Flake" column. At the same marked aperture, going from 2x to 5x makes the blur on the subject slightly smaller. At f/2.8 it drops from 5.6 µm to 4.5 µm, because object-side resolution depends on N(M+1)/M, and that ratio shrinks as M rises. Higher magnification makes diffraction look worse per pixel, but it doesn't erase detail on the crystal. Depth is the cost.
How Much Depth a Tilted Plate Really Needs
People obsess over plate thickness. Thin hexagonal plates are far shallower than 100 µm, so thickness is rarely the problem. Tilt is.
A plate resting on a mitten fibre or a thread is almost never perfectly square to the lens. The depth it spans is roughly diameter × sin(tilt angle), plus its own thickness.
| Tilt Angle | Depth Spanned, 2 mm Plate | Depth Spanned, 3 mm Plate |
|---|---|---|
| 1° | 35 µm | 52 µm |
| 2° | 70 µm | 105 µm |
| 3° | 105 µm | 157 µm |
| 5° | 174 µm | 261 µm |
Three degrees is a tilt you won't notice by eye, and it uses up more depth than the plate's thickness. I lost a whole series of stellar dendrites to this: the left arms were sharp and the right arms faded. I assumed I'd missed focus. I hadn't. The plate was sitting on a fibre that lifted one edge by about the width of a hair.
How to Square a Plate to the Sensor Before Shooting
- Use a substrate with a flat top surface. Dark wool looks great, but its fibres hold crystals at random angles. A smooth dark glass slide or a fine-weave fabric keeps plates flatter.
- Rack focus through the plate before you shoot. If opposite arms come into focus at different points, the plate is tilted.
- Nudge the plate, not the camera. Tilting the camera to match the plate also shifts the background and lighting angle. A soft brush works for repositioning.
How to Pick the Best Single-Frame Aperture for Any Plate
Here's the formula I use more than any other. Defocus blur at the edges of the depth shrinks as you stop down, and diffraction blur grows. The total is smallest when the two are equal:
N_eff (best) = M × √(D / 2.68 µm)
D is the total depth you need sharp, in microns. Divide the result by (M + 1) to get the marked aperture, assuming P = 1.
| Depth to Cover (D) | Best Marked f at 3x | Best Marked f at 5x | Smallest Blur Spot on the Flake |
|---|---|---|---|
| 30 µm | f/2.5 | f/2.8 | 4.5 µm |
| 60 µm | f/3.5 | f/3.9 | 6.3 µm |
| 100 µm | f/4.6 | f/5.1 | 8.2 µm |
| 150 µm | f/5.6 | f/6.2 | 10.0 µm |
Two results here are more useful than any calculator:
- The best marked aperture barely moves with magnification. For about 100 µm of depth, it sits between f/4 and f/5.6 whether you're at 2x or 5x. That's why f/4 at 3x works so well for a slightly tilted plate.
- The finest detail you can resolve on the crystal depends only on depth, not magnification. It equals √(1.22 × λ × D). For a 100 µm span, that's about 8 µm on the subject at any magnification. Going from 3x to 5x on a single frame doesn't add detail. It just spreads the same 8 µm blur over more pixels.
That second point changed how I shoot. If the plate needs 100 µm of depth and I'm taking a single frame, I don't shoot at 5x. I shoot at about 3x, frame wider, and keep the shutter speed I'd lose at 5x.
When to Stop Fighting Diffraction and Stack Instead
Stacking separates depth from aperture. Each frame only has to be sharp across a thin slice, so you can open up to the aperture with the smallest blur spot on the flake. Usually that's wide open or one stop down, depending on how well the lens performs at its widest setting.
How to Size Focus Steps for Snowflake Stacks
- Step at about half the DOF figure from the first table for your aperture and magnification. At f/2.8 and 5x (30 µm DOF), I step about 15 µm.
- Count frames from tilt, not thickness. A 2 mm plate tilted 3° spans about 105 µm. At 15 µm steps, that's 7 or 8 frames.
- Keep stacks short. Plates sublimate and lose fine edge detail while you work, even in ordinary winter cold. Fewer frames at a slightly smaller aperture often beat a long stack of a crystal that's changing shape.
How to Use Light Color to Shrink the Airy Disk
Airy diameter scales with wavelength. Moving from green (550 nm) to blue (450 nm) light makes it about 18% smaller. If you backlight plates through a colored filter, a blue or cyan gel gives a small but real resolution gain. A warm amber backlight costs you resolution, even if it looks nice.
How Output Size Changes the Diffraction Math
Pixel-peeping isn't the final test. If you're exporting at 2048 px wide from an 8192 px-wide 45 MP file, each output pixel covers about 17.6 µm of sensor. A 30 µm Airy disk at f/22.4 effective is then under two output pixels and looks clean.
- Web and social sharing: f/22 to f/24 effective is usable, so single frames at f/5.6 and 3x are fine.
- Large prints or 100% crops: Keep N_eff below about f/12 on a 45 MP sensor. At 3x and above, that's only possible with stacking.
- Higher-resolution sensors (smaller pixels, around 3.8 µm on 61 MP full frame) show diffraction earlier at 100%, but they don't make output at the same size any worse.
FAQ
Does a higher-megapixel sensor make diffraction worse?
No. It makes diffraction more visible at 100% view, but output at the same print or screen size is never worse. The Airy disk at a given effective aperture is the same physical size whatever the pixel count. Smaller pixels just sample it more finely.
Why does my camera show f/5.6 when the lens is set to f/2.8?
Your camera is reporting the effective aperture, not the marked one. At 1x with a pupil magnification near 1, f/2.8 marked becomes f/5.6 effective. Some bodies do this automatically with matched macro lenses, and some never do.
Can I trust the pupil magnification of 1 assumption?
Only for roughly symmetric lens designs. Telephoto-style macro lenses often have P well below 1, which makes the real effective aperture smaller (a higher f-number) than N × (M + 1). Measure the exposure drop with the grey-card method above if you need exact numbers.
Is stacking always better than a single frame?
No. When the plate is close to flat and the crystal is changing quickly, a single frame at the best aperture can beat a stack. A 30 µm span at 3x needs only about f/2.5 marked, and a single frame there has no stacking artifacts and no risk of the crystal changing between frames.
Do extension tubes change these numbers?
They change magnification, and the formulas work on magnification, so use your actual M. Tubes on a short lens can also shift the pupil ratio. Measure M by photographing a ruler and dividing the ruler width that fills the frame into your sensor width.
Does the hexagonal shape itself affect depth of field?
Not directly. The crystal's tilt and diameter set the depth you need, not its symmetry. Branched dendrites do expose tilt sooner than plain plates, though, because their thin tips make even small defocus easy to see.
What to Do Next
Before your next snowfall:
- Measure your real magnification on a ruler at each setting you plan to use, and write down the sensor width it covers.
- Run the grey-card exposure test at that magnification to confirm your effective aperture and pupil magnification.
- Make a pocket card with three numbers per magnification: N_eff at your two favorite marked apertures, and the best marked aperture for 50 µm and 100 µm depths, calculated with the formula above.
- Rack-check every plate for tilt before deciding between a single frame and a stack. That check decides more of your results than aperture choice does.
Then shoot one plate three ways: single frame at the calculated best aperture, single frame at f/8 marked, and a short stack wide open. Compare them at 100% and at your usual export size.
0 Comments